论文部分内容阅读
通过概率神经网络PNN对金融交易时间序列数据的预测偏移误差分类实现对交易异常与否的分类,并将其与前馈神经网络BP、后馈神经网络Elman、竞争型神经网络LVQ、SOM等4种经典类型的分类效率进行比较,结果发现PNN在相近预测精度前提下在网络结构、运行效率方面都有明显优势,适合金融交易海量数据的监测分析.
Through the classification of prediction error of financial transaction time series data by probabilistic neural network PNN, it classifies the transaction abnormalities or not, and compares it with feedforward neural network BP, feedforward neural network Elman, competitive neural network LVQ, SOM and so on. Four classic types of classification efficiency are compared. The results show that PNN has obvious advantages in network structure and operation efficiency under the premise of similar prediction accuracy, which is suitable for the monitoring and analysis of mass data of financial transactions.